AI agents struggle to run a real store
A new benchmark tests AI agents by having them run a live online shop, and no current model comes close to a simple scripted policy.
A new benchmark tests AI agents by having them run a live online shop, and no current model comes close to a simple scripted policy.